CN105117739A - Clothes classifying method based on convolutional neural network - Google Patents
Clothes classifying method based on convolutional neural network Download PDFInfo
- Publication number
- CN105117739A CN105117739A CN201510457010.2A CN201510457010A CN105117739A CN 105117739 A CN105117739 A CN 105117739A CN 201510457010 A CN201510457010 A CN 201510457010A CN 105117739 A CN105117739 A CN 105117739A
- Authority
- CN
- China
- Prior art keywords
- layer
- convolutional neural
- clothes
- neural networks
- training
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V30/00—Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
- G06V30/10—Character recognition
- G06V30/19—Recognition using electronic means
- G06V30/192—Recognition using electronic means using simultaneous comparisons or correlations of the image signals with a plurality of references
- G06V30/194—References adjustable by an adaptive method, e.g. learning
Abstract
The invention discloses a clothes classifying method based on a convolutional neural network. The method comprises the following steps of acquiring clothes image samples, and dividing the samples into training samples and testing samples; preprocessing the training samples and the testing samples; constructing a convolutional neural network model; performing training of two stages including a forward propagation stage and a backward propagation stage on the convolutional neural network model through preprocessed training samples, finishing the training when the error calculated during the backward propagation stage reaches a desired value, and acquiring a parameter of the convolutional neural network model; testing the preprocessed testing samples by using the trained convolutional neural network model and outputting final clothes classifying results. The convolutional neural network model can make clothes images directly serve as network inputs, extract image features in an implicit way and establish a global feature expression. Compared with a manually designed feature extraction way, the method is more convenient and accurate. The problem that a conventional algorithm leads to low clothes classifying accuracy is solved.
Description
Technical field
The present invention relates to a kind of clothes sorting technique based on convolutional neural networks (ConvolutionalNeuralNetworks, CNN), belong to technical field of image information processing.
Background technology
At present, researchers have proposed many algorithms realizing clothes automatic classification.The people such as Pan propose to use BP neural network recognization knitted fabric.The people such as Ben proposes based on text feature and support vector machine the recognition methods of knitted fabric.The people such as Liu propose based on attitude estimation and use color, clothes is divided into 23 classes by the features such as SIFT, HOG.The people such as Bourdev work out a system to describe the appearance image of people, and they employ 9 attribute, the characteristics such as the such as male sex, T-shirt, long hair.In addition, the segmentation for clothes is also study hotspot.The people such as Hu propose to use the prospect based on limited moral labor Triangle ID (CDT) and background estimating, and this method is without any need for predefined dress form.The people such as Weber then introduce a novel method, utilize gesture detector to remove to process the occlusion issue of clothes.Manfred proposes the clothes dividing method in popular shop database.Researchers are also devoted to the classification to clothes attribute, the attributes such as such as color, collar, coat-sleeve.Chen introduces a full automatic system, and this system can produce the clothes can naming attribute list.The people such as Lorenzo-Navarro have then done one group of experiment, and the object of experiment evaluates the ability of LBP and HOG descriptor in clothes attribute.
Clothes classification is the research topic very with development prospect, traditional sorting algorithm generally adopts two-stage process, the first step calculates the artificial feature arranged from input picture, and second step removes training sorter, for the classification of test data according to the feature calculated.Due to the limitation of the feature of engineer, the quality of classic method effect depends on to a great extent thinks that whether the feature selected is reasonable, and tool bears the character of much blindness, the problem that ubiquity classification accuracy is low.Therefore, mainly there are 2 limitation in current clothes sorting algorithm.The first, traditional feature can not reach satisfied classifying quality, especially to the classification of like attribute.The second, do not have disclosed garment data to carry out the existing algorithm of objective evaluation at present.
Summary of the invention
Technical matters to be solved by this invention is: provide a kind of clothes sorting technique based on convolutional neural networks, and design convolutional neural networks model, the feature of image in learning database, sets up the feature representation of the overall situation, improve clothes classification accuracy.
The present invention is for solving the problems of the technologies described above by the following technical solutions:
Based on a clothes sorting technique for convolutional neural networks, comprise the following steps:
Step 1, obtains image of clothing sample, image of clothing sample is divided into training sample and test sample book;
Step 2, carries out pre-service to training sample and test sample book;
Step 3, build convolutional neural networks model, this convolutional neural networks model comprises 5 layers of convolutional layer, 3 layers of down-sampling layer, 3 layers of full articulamentum;
Step 4, the pretreated training sample of step 2 is utilized to carry out the training in propagated forward and two stages of back-propagating to convolutional neural networks model described in step 3, when back-propagating trains the error calculated to reach expectation value, training terminates, and obtains the parameter of convolutional neural networks model;
Step 5, utilizes step 4 to train the convolutional neural networks model terminated to test the pretreated test sample book of step 2, and exports final clothes classification results.
Preferably, the process obtaining image of clothing sample described in step 1 is: clothes are divided into men's clothing 8 class, women's dress 8 class carries out image pattern acquisition, wherein men's clothing 8 class is jacket, shirt, wind coat, western-style clothes, charge garment, sweater, down jackets, T-shirt, and women's dress 8 class is cheongsam, shirt, wind coat, western-style clothes, one-piece dress, defend clothing, down jackets, T-shirt.
Preferably, carrying out pretreated process to training sample and test sample book described in step 2 is: the size of training sample and test sample book is adjusted to 120 × 120 pixels.
Preferably, the computing formula of convolutional layer described in step 3 is:
wherein,
for convolutional layer l
ca jth output map of layer, f is activation function, M
jfor the set that input feature vector maps, * is convolution operation,
for convolutional layer l
ca jth output map of layer and last layer i-th convolution kernel inputted between figure, 1≤i≤max (l
cin), max (l
cin) be l
cthe maximum number of layer input figure, 1≤i≤max (l
cout), max (l
cout) be l
cthe maximum number of layer output map,
for convolutional layer l
cthe additional deviation of a jth output map of layer, l
c=1 ..., 5.
Preferably, described in step 3, the computing formula of down-sampling layer is:
wherein,
for down-sampling layer l
sa jth output map of layer, f is activation function, and S is down-sampling function,
be respectively down-sampling layer l
sthe multiplier deviation of a jth output map of layer, additional deviation, l
s=1 ..., 3.
Preferably, the computing formula of error described in step 4 is:
wherein, n is the total sample number of training sample, and m is classification number, t be by network activation function f (x)=max (0, the matrix of m × 1 x) exported, t
labelfor the label of training sample, it is the two values matrix of m × 1.
The present invention adopts above technical scheme compared with prior art, has following technique effect:
1, the present invention is based on the clothes sorting technique of convolutional neural networks, the convolutional neural networks framework of design can using image of clothing directly as the input of network, implicitly the feature of image is extracted, set up the feature representation of the overall situation, convenient compared to the feature extraction of engineer and accurate.
2, the present invention is based on the clothes sorting technique of convolutional neural networks, solve the problem that existing algorithm is low to clothes classification accuracy.
Accompanying drawing explanation
Fig. 1 is the schematic flow sheet of the clothes sorting technique that the present invention is based on convolutional neural networks.
Fig. 2 is convolutional neural networks structural representation in the embodiment of the present invention.
Fig. 3 is each convolutional layer output characteristic schematic diagram in the embodiment of the present invention.
Fig. 4 is the classification results schematic diagram that sorting technique of the present invention contrasts other three kinds of distinct methods.
Fig. 5 is that sorting technique of the present invention and other three kinds of distinct methods are to the classification results schematic diagram of each class clothes.
Embodiment
Be described below in detail embodiments of the present invention, the example of described embodiment is shown in the drawings, and wherein same or similar label represents same or similar element or has element that is identical or similar functions from start to finish.Being exemplary below by the embodiment be described with reference to the drawings, only for explaining the present invention, and can not limitation of the present invention being interpreted as.
As shown in Figure 1, for the present invention is based on the schematic flow sheet of the clothes sorting technique of convolutional neural networks, according to Fig. 1, the following examples are described in detail.
Step 1, acquisition image of clothing, set up training sample and test sample book.According to garment common on market, clothes are divided into 16 classes, wherein men's clothing 8 class, women's dress 8 class, download picture corresponding to this attribute by related software and artificial mode from internet, create new database.Wherein men's clothing is specifically subdivided into the clothes of jacket, shirt, wind coat, western-style clothes, charge garment, sweater, down jackets, T-shirt eight attribute; Women's dress is specifically subdivided into cheongsam, shirt, wind coat, western-style clothes, one-piece dress, defends the clothes of clothing, down jackets, T-shirt eight attribute.Image of clothing sample size comes to 33965, and therefrom random selecting 27565 samples are as training sample, and 6400 samples are as test sample book, and test ratio is 18.84%.
Step 2, pre-service is carried out to image of clothing, image size in adjustment database.Unification is adjusted to 120 × 120 pixels image size, as the input of convolutional neural networks.
Step 3, design effective network structure and go to express complicated image of clothing from the suitable feature of study.Convolutional neural networks of the present invention 23 layers altogether, has only carried 5 layers of convolutional layer here, 3 layers of down-sampling layer, 3 layers of full articulamentum, and the last layer of convolutional layer may be convolutional layer, also may be down-sampling layer.As shown in Figure 2, picture comprises through the detailed process of every one deck:
Ground floor is that convolutional layer C1 is made up of 16 Feature Mapping, and size is 55 × 55;
Second layer S1 layer, to C1 layer down-sampling, comprises the Feature Mapping of 16 27 × 27;
Third layer C2 layer is also a convolutional layer, and picture size becomes 27 × 27;
4th layer of S2 layer is a down-sampling layer, and picture size becomes 13 × 13;
Layer 5 is convolutional layer C3, and picture size becomes 13 × 13;
Layer 6 is convolutional layer C4, and picture size is 13 × 13;
Layer 7 is convolutional layer C5, and picture size is 13 × 13;
8th layer of S5 layer carries out down-sampling to C5 layer, and picture size becomes 6 × 6;
9th layer and the tenth layer is full articulamentum fc6, fc7, exports the Feature Mapping of 100 1 × 1;
Eleventh floor is full articulamentum fc8, exports the proper vector of one 16 dimension, delivers to the classification of softmax layer, output category result.
Step 4, training convolutional neural networks model.The training need continuous print iteration optimization of CNN model, it can go according to Iterative classification result the parameter adjusting next iteration.Picture is input to network, and through propagated forward and two training stages of back-propagating, propagated forward process is a sample input network, calculates corresponding actual output; Back-propagating process calculates actual output and the desirable difference exported, and according to error rate, continues to optimize network parameter, carry out the training of model.
Step 5, test sample book to be input in the CNN model of having trained and to classify, assorting process only comprises propagated forward, and image is successively transmitted to output layer output category result through network.
As shown in Figure 3, be convolutional layer output characteristic schematic diagram each in embodiment, traditional feature extraction algorithm such as HOG feature can only describe the edge feature of picture, lack the overall situation feature interpretation and to noise-sensitive.Can the feature of abstract image layer by layer unlike, CNN model with HOG feature, the feature that CNN model exports can the global characteristics of Description Image, instead of the marginal information of low layer.Therefore, the feature of CNN study can express the high-level semantics features of image effectively, is effective method more.
As shown in Figure 4, the comparison of classification accuracy of CNN and HOG+SVM of the present invention, BOW, HSV+SVM is given.Transverse axis represents method, and the longitudinal axis represents accuracy rate, and as can be seen from the figure, the accuracy rate of the accuracy rate of CNN to be the accuracy rate of 75.50%, HOG+SVM be 60.36%, BOW is the accuracy rate of 56.27%, HSV+SVM is 20.58%.Can find out, the accuracy rate of CNN is all higher than additive method.
As shown in Figure 5, give three kinds of distinct methods comparing the classification accuracy of each class clothes of sorting technique of the present invention and other, on the whole, CNN also can embody certain advantage in the accuracy rate of every oneclass classification.Convolution can obtain the marginal information of good clothes image, study semantic feature.Therefore, the stronger clothes image of edge feature is as charge garment and cheongsam, and the accuracy rate of their classification is just higher.But for the clothes that style is similar, their edge feature is easily obscured, and therefore accuracy rate is also relatively low.
Relative to low layer or middle level features, algorithm of the present invention can learn global information and the semantic feature of image, can overcome the problem of the low accuracy rate of attributive classification, compare with the Feature Extraction Method of traditional engineer, algorithm of the present invention can present certain advantage.
Above embodiment is only and technological thought of the present invention is described, can not limit protection scope of the present invention with this, and every technological thought proposed according to the present invention, any change that technical scheme basis is done, all falls within scope.
Claims (6)
1., based on a clothes sorting technique for convolutional neural networks, it is characterized in that: comprise the following steps:
Step 1, obtains image of clothing sample, image of clothing sample is divided into training sample and test sample book;
Step 2, carries out pre-service to training sample and test sample book;
Step 3, build convolutional neural networks model, this convolutional neural networks model comprises 5 layers of convolutional layer, 3 layers of down-sampling layer, 3 layers of full articulamentum;
Step 4, the pretreated training sample of step 2 is utilized to carry out the training in propagated forward and two stages of back-propagating to convolutional neural networks model described in step 3, when back-propagating trains the error calculated to reach expectation value, training terminates, and obtains the parameter of convolutional neural networks model;
Step 5, utilizes step 4 to train the convolutional neural networks model terminated to test the pretreated test sample book of step 2, and exports final clothes classification results.
2. as claimed in claim 1 based on the clothes sorting technique of convolutional neural networks, it is characterized in that: the process obtaining image of clothing sample described in step 1 is: clothes are divided into men's clothing 8 class, women's dress 8 class carries out image pattern acquisition, wherein men's clothing 8 class is jacket, shirt, wind coat, western-style clothes, charge garment, sweater, down jackets, T-shirt, and women's dress 8 class is cheongsam, shirt, wind coat, western-style clothes, one-piece dress, defend clothing, down jackets, T-shirt.
3. as claimed in claim 1 based on the clothes sorting technique of convolutional neural networks, it is characterized in that: carrying out pretreated process to training sample and test sample book described in step 2 is: the size of training sample and test sample book is adjusted to 120 × 120 pixels.
4. as claimed in claim 1 based on the clothes sorting technique of convolutional neural networks, it is characterized in that: the computing formula of convolutional layer described in step 3 is:
Wherein,
for convolutional layer l
ca jth output map of layer, f is activation function, M
jfor the set that input feature vector maps, * is convolution operation,
for convolutional layer l
ca jth output map of layer and last layer i-th convolution kernel inputted between figure, 1≤i≤max (l
cin), max (l
cin) be l
cthe maximum number of layer input figure, 1≤i≤max (l
cout), max (l
cout) be l
cthe maximum number of layer output map,
for convolutional layer l
cthe additional deviation of a jth output map of layer, l
c=1 ..., 5.
5. as claimed in claim 1 based on the clothes sorting technique of convolutional neural networks, it is characterized in that: described in step 3, the computing formula of down-sampling layer is:
Wherein,
for down-sampling layer l
sa jth output map of layer, f is activation function, and S is down-sampling function,
be respectively down-sampling layer l
sthe multiplier deviation of a jth output map of layer, additional deviation, l
s=1 ..., 3.
6. as claimed in claim 1 based on the clothes sorting technique of convolutional neural networks, it is characterized in that: the computing formula of error described in step 4 is:
Wherein, n is the total sample number of training sample, and m is classification number, t be by network activation function f (x)=max (0, the matrix of m × 1 x) exported, t
labelfor the label of training sample, it is the two values matrix of m × 1.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510457010.2A CN105117739A (en) | 2015-07-29 | 2015-07-29 | Clothes classifying method based on convolutional neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510457010.2A CN105117739A (en) | 2015-07-29 | 2015-07-29 | Clothes classifying method based on convolutional neural network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN105117739A true CN105117739A (en) | 2015-12-02 |
Family
ID=54665722
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510457010.2A Pending CN105117739A (en) | 2015-07-29 | 2015-07-29 | Clothes classifying method based on convolutional neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105117739A (en) |
Cited By (26)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105447529A (en) * | 2015-12-30 | 2016-03-30 | 商汤集团有限公司 | Costume detection and attribute value identification method and system |
| CN105787490A (en) * | 2016-03-24 | 2016-07-20 | 南京新与力文化传播有限公司 | Commodity fashion identification method and device based on deep learning |
| CN105787510A (en) * | 2016-02-26 | 2016-07-20 | 华东理工大学 | System and method for realizing subway scene classification based on deep learning |
| CN105791913A (en) * | 2016-04-27 | 2016-07-20 | 青岛海信传媒网络技术有限公司 | Multi-user telephone configuration information setting method, multi-user telephone configuration information setting system and cloud server |
| CN106022375A (en) * | 2016-05-19 | 2016-10-12 | 东华大学 | HU invariant moment and support vector machine-based garment style identification method |
| CN106126579A (en) * | 2016-06-17 | 2016-11-16 | 北京市商汤科技开发有限公司 | Object identification method and device, data processing equipment and terminal unit |
| CN106203330A (en) * | 2016-07-08 | 2016-12-07 | 西安理工大学 | A kind of vehicle classification method based on convolutional neural networks |
| CN106875203A (en) * | 2015-12-14 | 2017-06-20 | 阿里巴巴集团控股有限公司 | A kind of method and device of the style information for determining commodity picture |
| CN106875061A (en) * | 2017-02-20 | 2017-06-20 | 佛山市络思讯科技有限公司 | Method and relevant apparatus that a kind of destination path determines |
| CN106874296A (en) * | 2015-12-14 | 2017-06-20 | 阿里巴巴集团控股有限公司 | A kind of style recognition methods of commodity and device |
| CN106934408A (en) * | 2015-12-29 | 2017-07-07 | 北京大唐高鸿数据网络技术有限公司 | Identity card picture sorting technique based on convolutional neural networks |
| CN107169508A (en) * | 2017-04-17 | 2017-09-15 | 杭州电子科技大学 | A kind of cheongsam Image emotional semantic method for recognizing semantics based on fusion feature |
| CN107330451A (en) * | 2017-06-16 | 2017-11-07 | 西交利物浦大学 | Clothes attribute retrieval method based on depth convolutional neural networks |
| CN107683469A (en) * | 2015-12-30 | 2018-02-09 | 中国科学院深圳先进技术研究院 | A kind of product classification method and device based on deep learning |
| CN107704882A (en) * | 2017-10-13 | 2018-02-16 | 上海工程技术大学 | A kind of kinds of laundry recognition methods and system based on digital image processing techniques |
| CN107886049A (en) * | 2017-10-16 | 2018-04-06 | 江苏省气象服务中心 | A kind of visibility identification method for early warning based on camera probe |
| CN108614884A (en) * | 2018-05-03 | 2018-10-02 | 桂林电子科技大学 | A kind of image of clothing search method based on convolutional neural networks |
| CN109271846A (en) * | 2018-08-01 | 2019-01-25 | 深圳云天励飞技术有限公司 | Personal identification method, apparatus and storage medium |
| CN109993213A (en) * | 2019-03-06 | 2019-07-09 | 上海工程技术大学 | A kind of automatic identifying method for garment elements figure |
| CN111415221A (en) * | 2020-03-19 | 2020-07-14 | 重庆邮电大学 | Interpretable convolutional neural network-based garment recommendation method and device and terminal |
| CN111444974A (en) * | 2020-04-02 | 2020-07-24 | 成都三零凯天通信实业有限公司 | Clothing classification method based on zero sample recognition |
| CN112164013A (en) * | 2020-10-14 | 2021-01-01 | 上海影卓信息科技有限公司 | Portrait reloading method, system and medium |
| CN112446767A (en) * | 2020-12-14 | 2021-03-05 | 武汉纺织大学 | Clothing recommendation system |
| CN112883163A (en) * | 2021-03-05 | 2021-06-01 | 南通大学 | Computer application method for garment process classification |
| CN114694012A (en) * | 2022-04-07 | 2022-07-01 | 哈尔滨理工大学 | Clothes identification method based on convolutional neural network |
| CN116797533A (en) * | 2023-03-24 | 2023-09-22 | 东莞市冠锦电子科技有限公司 | Appearance defect detection method and system for power adapter |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20110222724A1 (en) * | 2010-03-15 | 2011-09-15 | Nec Laboratories America, Inc. | Systems and methods for determining personal characteristics |
| CN103902987A (en) * | 2014-04-17 | 2014-07-02 | 福州大学 | Station caption identifying method based on convolutional network |
| CN103927534A (en) * | 2014-04-26 | 2014-07-16 | 无锡信捷电气股份有限公司 | Sprayed character online visual detection method based on convolutional neural network |
| CN104102919A (en) * | 2014-07-14 | 2014-10-15 | 同济大学 | Image classification method capable of effectively preventing convolutional neural network from being overfit |
| CN104504055A (en) * | 2014-12-19 | 2015-04-08 | 常州飞寻视讯信息科技有限公司 | Commodity similarity calculation method and commodity recommending system based on image similarity |
| CN104778464A (en) * | 2015-05-04 | 2015-07-15 | 中国科学院重庆绿色智能技术研究院 | Garment positioning and detecting method based on depth convolution nerve network |
-
2015
- 2015-07-29 CN CN201510457010.2A patent/CN105117739A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20110222724A1 (en) * | 2010-03-15 | 2011-09-15 | Nec Laboratories America, Inc. | Systems and methods for determining personal characteristics |
| CN103902987A (en) * | 2014-04-17 | 2014-07-02 | 福州大学 | Station caption identifying method based on convolutional network |
| CN103927534A (en) * | 2014-04-26 | 2014-07-16 | 无锡信捷电气股份有限公司 | Sprayed character online visual detection method based on convolutional neural network |
| CN104102919A (en) * | 2014-07-14 | 2014-10-15 | 同济大学 | Image classification method capable of effectively preventing convolutional neural network from being overfit |
| CN104504055A (en) * | 2014-12-19 | 2015-04-08 | 常州飞寻视讯信息科技有限公司 | Commodity similarity calculation method and commodity recommending system based on image similarity |
| CN104778464A (en) * | 2015-05-04 | 2015-07-15 | 中国科学院重庆绿色智能技术研究院 | Garment positioning and detecting method based on depth convolution nerve network |
Non-Patent Citations (1)
| Title |
|---|
| 王强: ""基于CNN的字符识别方法研究"", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (36)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106875203A (en) * | 2015-12-14 | 2017-06-20 | 阿里巴巴集团控股有限公司 | A kind of method and device of the style information for determining commodity picture |
| CN106874296A (en) * | 2015-12-14 | 2017-06-20 | 阿里巴巴集团控股有限公司 | A kind of style recognition methods of commodity and device |
| CN106874296B (en) * | 2015-12-14 | 2021-06-04 | 阿里巴巴集团控股有限公司 | Method and device for identifying style of commodity |
| CN106934408A (en) * | 2015-12-29 | 2017-07-07 | 北京大唐高鸿数据网络技术有限公司 | Identity card picture sorting technique based on convolutional neural networks |
| CN105447529B (en) * | 2015-12-30 | 2020-11-03 | 商汤集团有限公司 | Method and system for detecting clothes and identifying attribute value thereof |
| CN107683469A (en) * | 2015-12-30 | 2018-02-09 | 中国科学院深圳先进技术研究院 | A kind of product classification method and device based on deep learning |
| CN105447529A (en) * | 2015-12-30 | 2016-03-30 | 商汤集团有限公司 | Costume detection and attribute value identification method and system |
| CN105787510A (en) * | 2016-02-26 | 2016-07-20 | 华东理工大学 | System and method for realizing subway scene classification based on deep learning |
| CN105787490A (en) * | 2016-03-24 | 2016-07-20 | 南京新与力文化传播有限公司 | Commodity fashion identification method and device based on deep learning |
| CN105791913A (en) * | 2016-04-27 | 2016-07-20 | 青岛海信传媒网络技术有限公司 | Multi-user telephone configuration information setting method, multi-user telephone configuration information setting system and cloud server |
| CN106022375A (en) * | 2016-05-19 | 2016-10-12 | 东华大学 | HU invariant moment and support vector machine-based garment style identification method |
| CN106022375B (en) * | 2016-05-19 | 2019-07-26 | 东华大学 | A kind of clothes fashion recognition methods based on HU not bending moment and support vector machines |
| CN106126579A (en) * | 2016-06-17 | 2016-11-16 | 北京市商汤科技开发有限公司 | Object identification method and device, data processing equipment and terminal unit |
| CN106203330A (en) * | 2016-07-08 | 2016-12-07 | 西安理工大学 | A kind of vehicle classification method based on convolutional neural networks |
| CN106875061A (en) * | 2017-02-20 | 2017-06-20 | 佛山市络思讯科技有限公司 | Method and relevant apparatus that a kind of destination path determines |
| CN107169508A (en) * | 2017-04-17 | 2017-09-15 | 杭州电子科技大学 | A kind of cheongsam Image emotional semantic method for recognizing semantics based on fusion feature |
| CN107169508B (en) * | 2017-04-17 | 2020-08-25 | 杭州电子科技大学 | Cheong-sam image emotion semantic recognition method based on fusion features |
| CN107330451A (en) * | 2017-06-16 | 2017-11-07 | 西交利物浦大学 | Clothes attribute retrieval method based on depth convolutional neural networks |
| CN107330451B (en) * | 2017-06-16 | 2020-06-26 | 西交利物浦大学 | Clothing attribute retrieval method based on deep convolutional neural network |
| CN107704882A (en) * | 2017-10-13 | 2018-02-16 | 上海工程技术大学 | A kind of kinds of laundry recognition methods and system based on digital image processing techniques |
| CN107886049A (en) * | 2017-10-16 | 2018-04-06 | 江苏省气象服务中心 | A kind of visibility identification method for early warning based on camera probe |
| CN107886049B (en) * | 2017-10-16 | 2022-08-26 | 江苏省气象服务中心 | Visibility recognition early warning method based on camera probe |
| CN108614884A (en) * | 2018-05-03 | 2018-10-02 | 桂林电子科技大学 | A kind of image of clothing search method based on convolutional neural networks |
| CN109271846A (en) * | 2018-08-01 | 2019-01-25 | 深圳云天励飞技术有限公司 | Personal identification method, apparatus and storage medium |
| CN109993213A (en) * | 2019-03-06 | 2019-07-09 | 上海工程技术大学 | A kind of automatic identifying method for garment elements figure |
| CN111415221A (en) * | 2020-03-19 | 2020-07-14 | 重庆邮电大学 | Interpretable convolutional neural network-based garment recommendation method and device and terminal |
| CN111415221B (en) * | 2020-03-19 | 2023-04-07 | 重庆邮电大学 | Clothing recommendation method and device based on interpretable convolutional neural network and terminal |
| CN111444974A (en) * | 2020-04-02 | 2020-07-24 | 成都三零凯天通信实业有限公司 | Clothing classification method based on zero sample recognition |
| CN112164013A (en) * | 2020-10-14 | 2021-01-01 | 上海影卓信息科技有限公司 | Portrait reloading method, system and medium |
| CN112164013B (en) * | 2020-10-14 | 2023-04-18 | 上海影卓信息科技有限公司 | Portrait reloading method, system and medium |
| CN112446767A (en) * | 2020-12-14 | 2021-03-05 | 武汉纺织大学 | Clothing recommendation system |
| CN112883163A (en) * | 2021-03-05 | 2021-06-01 | 南通大学 | Computer application method for garment process classification |
| CN114694012A (en) * | 2022-04-07 | 2022-07-01 | 哈尔滨理工大学 | Clothes identification method based on convolutional neural network |
| CN114694012B (en) * | 2022-04-07 | 2023-10-20 | 哈尔滨理工大学 | Clothing identification method based on convolutional neural network |
| CN116797533A (en) * | 2023-03-24 | 2023-09-22 | 东莞市冠锦电子科技有限公司 | Appearance defect detection method and system for power adapter |
| CN116797533B (en) * | 2023-03-24 | 2024-01-23 | 东莞市冠锦电子科技有限公司 | Appearance defect detection method and system for power adapter |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105117739A (en) | Clothes classifying method based on convolutional neural network | |
| Jiao et al. | AF-RCNN: An anchor-free convolutional neural network for multi-categories agricultural pest detection | |
| CN109685115B (en) | Fine-grained conceptual model with bilinear feature fusion and learning method | |
| CN109344825A (en) | A kind of licence plate recognition method based on convolutional neural networks | |
| Liu et al. | Matching-cnn meets knn: Quasi-parametric human parsing | |
| CN103106265B (en) | Similar image sorting technique and system | |
| CN107239731A (en) | A kind of gestures detection and recognition methods based on Faster R CNN | |
| CN104598885B (en) | The detection of word label and localization method in street view image | |
| CN105825502B (en) | A kind of Weakly supervised method for analyzing image of the dictionary study based on conspicuousness guidance | |
| CN109101938B (en) | Multi-label age estimation method based on convolutional neural network | |
| CN108009509A (en) | Vehicle target detection method | |
| CN109325398A (en) | A kind of face character analysis method based on transfer learning | |
| CN106919951A (en) | A kind of Weakly supervised bilinearity deep learning method merged with vision based on click | |
| CN103984948B (en) | A kind of soft double-deck age estimation method based on facial image fusion feature | |
| CN104866524A (en) | Fine classification method for commodity images | |
| CN105809123A (en) | Face detecting method and device | |
| CN110633708A (en) | Deep network significance detection method based on global model and local optimization | |
| CN104834941A (en) | Offline handwriting recognition method of sparse autoencoder based on computer input | |
| CN106022343A (en) | Fourier descriptor and BP neural network-based garment style identification method | |
| CN106022375A (en) | HU invariant moment and support vector machine-based garment style identification method | |
| CN106157375A (en) | A kind of threedimensional model component categories automatic marking method | |
| CN109002755A (en) | Age estimation model building method and estimation method based on facial image | |
| CN110222712B (en) | Multi-special-item target detection algorithm based on deep learning | |
| CN105138975B (en) | A kind of area of skin color of human body dividing method based on degree of depth conviction network | |
| CN106056165A (en) | Saliency detection method based on super-pixel relevance enhancing Adaboost classification learning |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20151202 |
|
| WD01 | Invention patent application deemed withdrawn after publication |